| Message ID | 1621406386-18838-3-git-send-email-chunfeng.yun@mediatek.com |
|---|---|
| State | New |
| Headers | show |
| Series | [v2,1/3] usb: common: usb-conn-gpio: fix NULL pointer dereference of charger | expand |
On Wed, May 19, 2021 at 02:39:46PM +0800, Chunfeng Yun wrote: > Vbus is already an optional supply, if the vbus-supply is not > provided in DTS, will use a dummy regulator, That statement is not entirely correct. The dummy regulator is substituted only if the supply is in fact not optional. The idea behind that is to allow DTS files that don't specify all required regulators to get away with it, based on the assumption that the supply is one of those always-on supplies that are often not described in DTS. > the warning log is as below: > "supply vbus not found, using dummy regulator" And the reason why we get that warning is to point out that the DTS has a bug and that it should be fixed (by adding a proper regulator to take the place of the dummy). > This reverts commit 4ddf1ac79e5f082451cd549283d2eb7559ab6ca9. But if you read the description of that commit, the purpose of that patch was in fact to make the supply completely optional in the case where we already have the VBUS supply specified for the USB port that the connector is parented to. So in that case the DTS doesn't have the bug because the VBUS supply is already specified for the USB port and therefore it doesn't have to be specified in the USB connector again. In fact, specifying it twice can lead to a situation where the USB port may not be able to switch the VBUS supply on or off because the setting might conflict with that of the USB connector. So unless there's a real reason why this is needed, I don't think this should be applied. Thierry > Cc: Thierry Reding <treding@nvidia.com> > Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com> > --- > v2: remove unused variable "need_vbus" > --- > drivers/usb/common/usb-conn-gpio.c | 30 ++++++------------------------ > 1 file changed, 6 insertions(+), 24 deletions(-) > > diff --git a/drivers/usb/common/usb-conn-gpio.c b/drivers/usb/common/usb-conn-gpio.c > index dfbbc4f51ed6..65d89140cd19 100644 > --- a/drivers/usb/common/usb-conn-gpio.c > +++ b/drivers/usb/common/usb-conn-gpio.c > @@ -91,14 +91,14 @@ static void usb_conn_detect_cable(struct work_struct *work) > return; > } > > - if (info->last_role == USB_ROLE_HOST && info->vbus) > + if (info->last_role == USB_ROLE_HOST) > regulator_disable(info->vbus); > > ret = usb_role_switch_set_role(info->role_sw, role); > if (ret) > dev_err(info->dev, "failed to set role: %d\n", ret); > > - if (role == USB_ROLE_HOST && info->vbus) { > + if (role == USB_ROLE_HOST) { > ret = regulator_enable(info->vbus); > if (ret) > dev_err(info->dev, "enable vbus regulator failed\n"); > @@ -106,9 +106,8 @@ static void usb_conn_detect_cable(struct work_struct *work) > > info->last_role = role; > > - if (info->vbus) > - dev_dbg(info->dev, "vbus regulator is %s\n", > - regulator_is_enabled(info->vbus) ? "enabled" : "disabled"); > + dev_dbg(info->dev, "vbus regulator is %s\n", > + regulator_is_enabled(info->vbus) ? "enabled" : "disabled"); > > power_supply_changed(info->charger); > } > @@ -175,7 +174,6 @@ static int usb_conn_probe(struct platform_device *pdev) > { > struct device *dev = &pdev->dev; > struct usb_conn_info *info; > - bool need_vbus = true; > int ret = 0; > > info = devm_kzalloc(dev, sizeof(*info), GFP_KERNEL); > @@ -205,23 +203,7 @@ static int usb_conn_probe(struct platform_device *pdev) > > INIT_DELAYED_WORK(&info->dw_det, usb_conn_detect_cable); > > - /* > - * If the USB connector is a child of a USB port and that port already provides the VBUS > - * supply, there's no need for the USB connector to provide it again. > - */ > - if (dev->parent && dev->parent->of_node) { > - if (of_find_property(dev->parent->of_node, "vbus-supply", NULL)) > - need_vbus = false; > - } > - > - if (!need_vbus) { > - info->vbus = devm_regulator_get_optional(dev, "vbus"); > - if (PTR_ERR(info->vbus) == -ENODEV) > - info->vbus = NULL; > - } else { > - info->vbus = devm_regulator_get(dev, "vbus"); > - } > - > + info->vbus = devm_regulator_get(dev, "vbus"); > if (IS_ERR(info->vbus)) { > ret = PTR_ERR(info->vbus); > return dev_err_probe(dev, ret, "failed to get vbus :%d\n", ret); > @@ -288,7 +270,7 @@ static int usb_conn_remove(struct platform_device *pdev) > > cancel_delayed_work_sync(&info->dw_det); > > - if (info->last_role == USB_ROLE_HOST && info->vbus) > + if (info->last_role == USB_ROLE_HOST) > regulator_disable(info->vbus); > > usb_role_switch_put(info->role_sw); > -- > 2.18.0 >
On Fri, May 21, 2021 at 03:20:23PM +0200, Thierry Reding wrote: > On Wed, May 19, 2021 at 02:39:46PM +0800, Chunfeng Yun wrote: > > Vbus is already an optional supply, if the vbus-supply is not > > provided in DTS, will use a dummy regulator, > > That statement is not entirely correct. The dummy regulator is > substituted only if the supply is in fact not optional. The idea behind > that is to allow DTS files that don't specify all required regulators to > get away with it, based on the assumption that the supply is one of > those always-on supplies that are often not described in DTS. > > > the warning log is as below: > > "supply vbus not found, using dummy regulator" > > And the reason why we get that warning is to point out that the DTS has > a bug and that it should be fixed (by adding a proper regulator to take > the place of the dummy). > > > This reverts commit 4ddf1ac79e5f082451cd549283d2eb7559ab6ca9. > > But if you read the description of that commit, the purpose of that > patch was in fact to make the supply completely optional in the case > where we already have the VBUS supply specified for the USB port that > the connector is parented to. > > So in that case the DTS doesn't have the bug because the VBUS supply is > already specified for the USB port and therefore it doesn't have to be > specified in the USB connector again. In fact, specifying it twice can > lead to a situation where the USB port may not be able to switch the > VBUS supply on or off because the setting might conflict with that of > the USB connector. > > So unless there's a real reason why this is needed, I don't think this > should be applied. I've dropped this from my tree now, thanks. greg k-h
On Fri, 2021-05-21 at 15:20 +0200, Thierry Reding wrote: > On Wed, May 19, 2021 at 02:39:46PM +0800, Chunfeng Yun wrote: > > Vbus is already an optional supply, if the vbus-supply is not > > provided in DTS, will use a dummy regulator, > > That statement is not entirely correct. The dummy regulator is > substituted only if the supply is in fact not optional. The idea behind > that is to allow DTS files that don't specify all required regulators to > get away with it, based on the assumption that the supply is one of > those always-on supplies that are often not described in DTS. Yes, you are right. But from the point of result, it indeed can help to handle the absent regulator. > > > the warning log is as below: > > "supply vbus not found, using dummy regulator" > > And the reason why we get that warning is to point out that the DTS has > a bug and that it should be fixed (by adding a proper regulator to take > the place of the dummy). > > > This reverts commit 4ddf1ac79e5f082451cd549283d2eb7559ab6ca9. > > But if you read the description of that commit, the purpose of that > patch was in fact to make the supply completely optional in the case > where we already have the VBUS supply specified for the USB port that > the connector is parented to. Could you please give an example you mentioned? It seems prefer to provide vbus supply in connector instead of port according to dt-binding Thanks > > So in that case the DTS doesn't have the bug because the VBUS supply is > already specified for the USB port and therefore it doesn't have to be > specified in the USB connector again. In fact, specifying it twice can > lead to a situation where the USB port may not be able to switch the > VBUS supply on or off because the setting might conflict with that of > the USB connector. > > So unless there's a real reason why this is needed, I don't think this > should be applied. > > Thierry > > > Cc: Thierry Reding <treding@nvidia.com> > > Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com> > > --- > > v2: remove unused variable "need_vbus" > > --- > > drivers/usb/common/usb-conn-gpio.c | 30 ++++++------------------------ > > 1 file changed, 6 insertions(+), 24 deletions(-) > > > > diff --git a/drivers/usb/common/usb-conn-gpio.c b/drivers/usb/common/usb-conn-gpio.c > > index dfbbc4f51ed6..65d89140cd19 100644 > > --- a/drivers/usb/common/usb-conn-gpio.c > > +++ b/drivers/usb/common/usb-conn-gpio.c > > @@ -91,14 +91,14 @@ static void usb_conn_detect_cable(struct work_struct *work) > > return; > > } > > > > - if (info->last_role == USB_ROLE_HOST && info->vbus) > > + if (info->last_role == USB_ROLE_HOST) > > regulator_disable(info->vbus); > > > > ret = usb_role_switch_set_role(info->role_sw, role); > > if (ret) > > dev_err(info->dev, "failed to set role: %d\n", ret); > > > > - if (role == USB_ROLE_HOST && info->vbus) { > > + if (role == USB_ROLE_HOST) { > > ret = regulator_enable(info->vbus); > > if (ret) > > dev_err(info->dev, "enable vbus regulator failed\n"); > > @@ -106,9 +106,8 @@ static void usb_conn_detect_cable(struct work_struct *work) > > > > info->last_role = role; > > > > - if (info->vbus) > > - dev_dbg(info->dev, "vbus regulator is %s\n", > > - regulator_is_enabled(info->vbus) ? "enabled" : "disabled"); > > + dev_dbg(info->dev, "vbus regulator is %s\n", > > + regulator_is_enabled(info->vbus) ? "enabled" : "disabled"); > > > > power_supply_changed(info->charger); > > } > > @@ -175,7 +174,6 @@ static int usb_conn_probe(struct platform_device *pdev) > > { > > struct device *dev = &pdev->dev; > > struct usb_conn_info *info; > > - bool need_vbus = true; > > int ret = 0; > > > > info = devm_kzalloc(dev, sizeof(*info), GFP_KERNEL); > > @@ -205,23 +203,7 @@ static int usb_conn_probe(struct platform_device *pdev) > > > > INIT_DELAYED_WORK(&info->dw_det, usb_conn_detect_cable); > > > > - /* > > - * If the USB connector is a child of a USB port and that port already provides the VBUS > > - * supply, there's no need for the USB connector to provide it again. > > - */ > > - if (dev->parent && dev->parent->of_node) { > > - if (of_find_property(dev->parent->of_node, "vbus-supply", NULL)) > > - need_vbus = false; > > - } > > - > > - if (!need_vbus) { > > - info->vbus = devm_regulator_get_optional(dev, "vbus"); > > - if (PTR_ERR(info->vbus) == -ENODEV) > > - info->vbus = NULL; > > - } else { > > - info->vbus = devm_regulator_get(dev, "vbus"); > > - } > > - > > + info->vbus = devm_regulator_get(dev, "vbus"); > > if (IS_ERR(info->vbus)) { > > ret = PTR_ERR(info->vbus); > > return dev_err_probe(dev, ret, "failed to get vbus :%d\n", ret); > > @@ -288,7 +270,7 @@ static int usb_conn_remove(struct platform_device *pdev) > > > > cancel_delayed_work_sync(&info->dw_det); > > > > - if (info->last_role == USB_ROLE_HOST && info->vbus) > > + if (info->last_role == USB_ROLE_HOST) > > regulator_disable(info->vbus); > > > > usb_role_switch_put(info->role_sw); > > -- > > 2.18.0 > >
On Mon, May 24, 2021 at 01:51:51PM +0800, Chunfeng Yun wrote: > On Fri, 2021-05-21 at 15:20 +0200, Thierry Reding wrote: > > On Wed, May 19, 2021 at 02:39:46PM +0800, Chunfeng Yun wrote: > > > Vbus is already an optional supply, if the vbus-supply is not > > > provided in DTS, will use a dummy regulator, > > > > That statement is not entirely correct. The dummy regulator is > > substituted only if the supply is in fact not optional. The idea behind > > that is to allow DTS files that don't specify all required regulators to > > get away with it, based on the assumption that the supply is one of > > those always-on supplies that are often not described in DTS. > Yes, you are right. > But from the point of result, it indeed can help to handle the absent > regulator. > > > > > the warning log is as below: > > > "supply vbus not found, using dummy regulator" > > > > And the reason why we get that warning is to point out that the DTS has > > a bug and that it should be fixed (by adding a proper regulator to take > > the place of the dummy). > > > > > This reverts commit 4ddf1ac79e5f082451cd549283d2eb7559ab6ca9. > > > > But if you read the description of that commit, the purpose of that > > patch was in fact to make the supply completely optional in the case > > where we already have the VBUS supply specified for the USB port that > > the connector is parented to. > Could you please give an example you mentioned? You can find examples of this in these: arch/arm64/boot/dts/nvidia/tegra210-p2597.dtsi arch/arm64/boot/dts/nvidia/tegra210-p3450-0000.dts arch/arm64/boot/dts/nvidia/tegra186-p2771-0000.dts > It seems prefer to provide vbus supply in connector instead of port > according to dt-binding My recollection is that the above (or at least some of them) predate USB connectors. It's possible that we could convert the above to have the VBUS supply listed in the connector instead of the port. However, since we have to preserve backwards compatibility with older device trees, we can't revert the commit anyway. Thierry
On Tue, 2021-05-25 at 13:36 +0200, Thierry Reding wrote: > On Mon, May 24, 2021 at 01:51:51PM +0800, Chunfeng Yun wrote: > > On Fri, 2021-05-21 at 15:20 +0200, Thierry Reding wrote: > > > On Wed, May 19, 2021 at 02:39:46PM +0800, Chunfeng Yun wrote: > > > > Vbus is already an optional supply, if the vbus-supply is not > > > > provided in DTS, will use a dummy regulator, > > > > > > That statement is not entirely correct. The dummy regulator is > > > substituted only if the supply is in fact not optional. The idea behind > > > that is to allow DTS files that don't specify all required regulators to > > > get away with it, based on the assumption that the supply is one of > > > those always-on supplies that are often not described in DTS. > > Yes, you are right. > > But from the point of result, it indeed can help to handle the absent > > regulator. > > > > > > > the warning log is as below: > > > > "supply vbus not found, using dummy regulator" > > > > > > And the reason why we get that warning is to point out that the DTS has > > > a bug and that it should be fixed (by adding a proper regulator to take > > > the place of the dummy). > > > > > > > This reverts commit 4ddf1ac79e5f082451cd549283d2eb7559ab6ca9. > > > > > > But if you read the description of that commit, the purpose of that > > > patch was in fact to make the supply completely optional in the case > > > where we already have the VBUS supply specified for the USB port that > > > the connector is parented to. > > Could you please give an example you mentioned? > > You can find examples of this in these: > > arch/arm64/boot/dts/nvidia/tegra210-p2597.dtsi > arch/arm64/boot/dts/nvidia/tegra210-p3450-0000.dts > arch/arm64/boot/dts/nvidia/tegra186-p2771-0000.dts > > > It seems prefer to provide vbus supply in connector instead of port > > according to dt-binding > > My recollection is that the above (or at least some of them) predate USB > connectors. > > It's possible that we could convert the above to have the VBUS supply > listed in the connector instead of the port. However, since we have to > preserve backwards compatibility with older device trees, we can't > revert the commit anyway. Got it, thanks a lot > > Thierry
diff --git a/drivers/usb/common/usb-conn-gpio.c b/drivers/usb/common/usb-conn-gpio.c index dfbbc4f51ed6..65d89140cd19 100644 --- a/drivers/usb/common/usb-conn-gpio.c +++ b/drivers/usb/common/usb-conn-gpio.c @@ -91,14 +91,14 @@ static void usb_conn_detect_cable(struct work_struct *work) return; } - if (info->last_role == USB_ROLE_HOST && info->vbus) + if (info->last_role == USB_ROLE_HOST) regulator_disable(info->vbus); ret = usb_role_switch_set_role(info->role_sw, role); if (ret) dev_err(info->dev, "failed to set role: %d\n", ret); - if (role == USB_ROLE_HOST && info->vbus) { + if (role == USB_ROLE_HOST) { ret = regulator_enable(info->vbus); if (ret) dev_err(info->dev, "enable vbus regulator failed\n"); @@ -106,9 +106,8 @@ static void usb_conn_detect_cable(struct work_struct *work) info->last_role = role; - if (info->vbus) - dev_dbg(info->dev, "vbus regulator is %s\n", - regulator_is_enabled(info->vbus) ? "enabled" : "disabled"); + dev_dbg(info->dev, "vbus regulator is %s\n", + regulator_is_enabled(info->vbus) ? "enabled" : "disabled"); power_supply_changed(info->charger); } @@ -175,7 +174,6 @@ static int usb_conn_probe(struct platform_device *pdev) { struct device *dev = &pdev->dev; struct usb_conn_info *info; - bool need_vbus = true; int ret = 0; info = devm_kzalloc(dev, sizeof(*info), GFP_KERNEL); @@ -205,23 +203,7 @@ static int usb_conn_probe(struct platform_device *pdev) INIT_DELAYED_WORK(&info->dw_det, usb_conn_detect_cable); - /* - * If the USB connector is a child of a USB port and that port already provides the VBUS - * supply, there's no need for the USB connector to provide it again. - */ - if (dev->parent && dev->parent->of_node) { - if (of_find_property(dev->parent->of_node, "vbus-supply", NULL)) - need_vbus = false; - } - - if (!need_vbus) { - info->vbus = devm_regulator_get_optional(dev, "vbus"); - if (PTR_ERR(info->vbus) == -ENODEV) - info->vbus = NULL; - } else { - info->vbus = devm_regulator_get(dev, "vbus"); - } - + info->vbus = devm_regulator_get(dev, "vbus"); if (IS_ERR(info->vbus)) { ret = PTR_ERR(info->vbus); return dev_err_probe(dev, ret, "failed to get vbus :%d\n", ret); @@ -288,7 +270,7 @@ static int usb_conn_remove(struct platform_device *pdev) cancel_delayed_work_sync(&info->dw_det); - if (info->last_role == USB_ROLE_HOST && info->vbus) + if (info->last_role == USB_ROLE_HOST) regulator_disable(info->vbus); usb_role_switch_put(info->role_sw);
Vbus is already an optional supply, if the vbus-supply is not provided in DTS, will use a dummy regulator, the warning log is as below: "supply vbus not found, using dummy regulator" This reverts commit 4ddf1ac79e5f082451cd549283d2eb7559ab6ca9. Cc: Thierry Reding <treding@nvidia.com> Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com> --- v2: remove unused variable "need_vbus" --- drivers/usb/common/usb-conn-gpio.c | 30 ++++++------------------------ 1 file changed, 6 insertions(+), 24 deletions(-)